Application of genetic algorithms and neural networks to predict the state change and emergency modes in aeroengines

The paper dwells on the problem of adaptability of digital automatic control systems to changes in the state and emergency modes of aeroengines in an operating process. The problem of additional data obtaining about the aeroengines state is solved by algorithmic methods using its built-in on-board mathematical model working in real-time mode. Two problems are considered.
The first is estimation of a change in the engine state due to the stochasticity of its characteristics. It is solved by parametric identification of the gas-path on the fundamental diagnostic matrix basis. Since the matrix determinant is close to zero, an ill-conditioned system of equations has gotten that has an infinite number of solutions. It is proposed to obtain an approximate solution in a given area using optimization based on a genetic algorithm. The proposed method has many times greater speed and greater accuracy of the solution than a similar solution obtained based on the numerical Monte Carlo method.
The second considered problem is diagnostics of the lean blow out (LBO) mode in the dry low emissions (DLE) combustion chamber. The problem is solved based on developing a neural network model of pressure pulsations that provides the given accuracy.